Topic 05 · Access & equity · Deep dive

Two-thirds of the world can't get scanned — and the other third waits in line

Equipment is only half the access story; someone has to read the images. The radiologist workforce runs from roughly 100 per million people in high-income countries to fewer than two per million in low-income ones, with fourteen African countries reporting none at all. But scarcity is not only a poor-country problem: inside wealthy universal systems, patients now wait months for an MRI. In May 2025 the World Health Assembly made imaging access, for the first time, a formal global-health target.

2/3
of the world's population has little or no access to medical imaging — the headline figure behind the 2025 WHO resolution
ISR / WHA78, 2025
14
African countries reported zero radiologists; low-income countries field ~2 per million vs ~98 per million in high-income ones
Lancet Oncology Commission 2021
18.1 wk
median wait for an MRI in Canada in 2025 (up from 16.2 in 2024) — a rich, universal system rationing by queue
Fraser Institute 2025
$179
lifetime productivity return per $1 invested in scaling up cancer imaging in LMICs (modelled, 2020–30)
Lancet Oncology Commission 2021
The workforce gradient

Machines you can buy; radiologists take a decade to make

Topic 03 mapped the hardware gap. This is the harder one to close: a scanner can be installed in months, but a radiologist takes years to train and is easily lured abroad. The result is a workforce gradient even steeper than the equipment one.

Radiologists per million people

Country and income-group estimates, latest available.
European Union average127/M · Sweden 270, Bulgaria 51
127
United States~1 per 10,000 people
~100
High-income averageLancet Commission
97.9
South Africa~1 per 100,000
~10
Ghana93 radiologists nationwide
3
Low-income average794 serve 697M people
<2
Sources: EU-REST study, Insights into Imaging 2025 (60,771 radiologists; EU average 127/M; range 51/M Bulgaria to 270/M Sweden); Lancet Oncology Commission 2021 (income-group densities; HIC 97.9/M); Collective Minds analysis (global mean ~45/M); Ghana study, Chin J Acad Radiol 2023; South Africa via RSSA/World Health Expo 2024 (~1 per 100,000). A peer-reviewed HPCSA workforce study puts South Africa closer to 17/M and the US at 100/M — definitions and years differ, so treat all as orders of magnitude, not precise counts.

The African extremes

People served per radiologist — the ratio, not the density.
Malawi1 radiologist per 8.8M people
8.8M
Nigeria~2.31 radiologists per million
~0.43M
Tunisiabest-resourced in Africa
23k
United Statesfor contrast
10k
Sources: RSNA Daily Bulletin (DeStigter, Malawi 1:8.8M & Tunisia 1:23,000, verbatim); Nigeria density (2.31/M ≈ 1:433,000) via J Glob Health Reports 2025. Fourteen African countries reported no radiologists at all (Lancet Commission). Bars show people-per-radiologist; Malawi's 8.8M is the single starkest workforce ratio recorded on the continent. Nigeria's absolute count is contested across sources (from <60 in older anecdotes to several hundred), so the per-capita density is the more reliable figure.
The brain-drain tax
South Africa qualifies ~20 radiologists a year — replacement level, not enough for growth (RSSA). Its graduates are actively recruited abroad: South Africa was the 2nd-largest source of Australia's overseas-trained radiologists in 2020 (5.7%, behind India's 7.8%).
Density hides geography
National averages mask internal deserts. In São Paulo, Brazil there are 10 radiologists per 100,000; in the country's north, 3. Ghana's radiologists cluster in three of sixteen regions. The rural poor lose twice.
Radiographers carry the load
In South Africa diagnostic radiographers make up 80%+ of the imaging workforce. Where radiologists are absent, task-shifting to radiographers and teleradiology (Topic 04) is not optional — it is the system.
Rationing inside rich systems

Wealthy, universal — and still waiting months for a scan

Access is not simply rich-versus-poor. Universal systems with plenty of machines still ration imaging by time, because supply of scanners and readers can't match demand. Canada, which surveys the queue annually, is the clearest window into how a rich country's access gap actually feels to a patient.

Median diagnostic wait times in Canada, 2025 (weeks)

Referral-to-scan, national medians, with the widest province spread noted.
MRI · national medianPEI reaches 52 weeks
18.1
CT · national medianOntario 6.0, Alberta 15.0
8.8
Ultrasound · national medianup to 14 in Newfoundland
5.4
PEI · MRI (worst province)a rich country's floor
52.0
Recommended · semi-urgent~30 days ≈ 4.3 weeks
4.3
Source: Fraser Institute, Waiting Your Turn 2025 (physician survey, Jan–May 2025, 13.1% response rate). MRI and CT waits both rose year-on-year (from 16.2 and 8.1 weeks in 2024). The gap between the 18.1-week MRI median and the ~4.3-week semi-urgent benchmark is the access failure — inside a G7 country with more machines than most of the world will see this decade.
Equity within borders

Same country, same insurance, different odds

Even where imaging is abundant, who reaches it — and how fast the follow-up comes — tracks race, income and rurality. US breast screening is the most-studied case: screening cut breast-cancer mortality ~40% since 1990, but the benefit did not land evenly.

US breast-cancer disparities

Where the imaging benefit fails to reach.
Black vs White mortalitydespite ~equal incidence
+41%
Under-50 Black vs White~double the mortality rate
~+100%
Same-day follow-up gapBlack vs White after abnormal mammogram
−50%
Mortality fall since 1990the benefit being unequally shared
−40%
Sources: SEER via RadioGraphics 2023 (40.7% higher Black mortality, ~6% lower incidence — the ~+41% gap); NEJM 2024 (40% overall mortality fall since 1990; 27.7 vs 20.0 deaths/100,000); ACS/BCRF 2024–25 & PMC (young Black women ~double, up to +111%, the mortality of young white women); breastcancer.org 2025 (Black women ~50% less likely to get same-day follow-up imaging/biopsy). The 40% mortality drop and the ~41% racial gap are the same statistic seen from two ends. Figures are observational and multi-causal (biology, treatment, insurance), not attributable to imaging access alone.

Geography of a screening desert

Delaware catchment study — facilities vs where Black & rural women live.
Fewer mammography unitsper SD more Black women in a tract
−68%
Fewer excellence centresBICOEs, same measure
−89%
Rural share of population41% of people…
41%
…but rural share of centres…only 22% of BICOEs
22%
Source: Delaware catchment & location-allocation analysis, PMC 2023 (a state chosen as broadly representative of US race and urban–rural mix). Facilities cluster away from the populations with the worst outcomes — a within-country access map that mirrors the global one.
The investment case

The cheapest life-saving intervention nobody funds

The Lancet Oncology Commission put an economic value on closing the gap. For eleven cancers modelled over 2020–30, scaling up imaging alone is startlingly cheap relative to its return — the reason imaging access finally reached the World Health Assembly floor.

Scale-up scenarioCost 2020–30Deaths avertedReturn on investment
Imaging alone$6.84B2.46M (3.2%)$179 lifetime productivity per $1; 54.9M life-years saved
Imaging alone (conservative human-capital)$6.84B2.46M$31.61 per $1; net benefit $209B
Imaging + treatment + quality of care9.55M (12.5%)$12.43 per $1; 232.3M life-years; net benefit $2.66T
Comprehensive (conservative human-capital)9.55M$2.46 per $1; net benefit $340B

Modelled figures from a Harvard microsimulation of 11 cancers (Lancet Oncology Commission, 2021). These are projections, not realised outcomes, and returns vary with the accounting method (full-income vs human-capital). All are scale-up-from-2020 scenarios; actual progress since has been partial.

A world first, in 2025
On 24 May 2025 the 78th World Health Assembly adopted WHA78.13, "Strengthening medical imaging capacity" — introduced by Cameroon, co-sponsored by Armenia, Brazil and Burkina Faso, and adopted by consensus. Imaging is now a named global-health target.
The funding mismatch
About 80% of cancer disability-adjusted life-years lost are in LMICs, yet only about 5% of global cancer-control funding is spent there. The imaging gap is a spending-priority gap first.
Where AI actually fits
The tools of Topic 04 matter most exactly here — triage and worklist-prioritisation for systems with too few readers. But AI needs machines, connectivity and a radiologist of last resort; it narrows the gap, it doesn't erase it.

Access has two failure modes, not one

The instinctive picture of imaging inequity — rich countries have scanners, poor countries don't — is correct but incomplete. There are two distinct failures. The first is absolute scarcity: fourteen African countries with no radiologist at all, Malawi with one per 8.8 million people, low-income countries fielding fewer than two readers per million against roughly a hundred in the United States. The second is rationing inside abundance: Canada, a G7 country, making patients wait a median 18.1 weeks for an MRI and up to a year in its worst province. Both are access failures; they just present differently. One is a queue that doesn't exist because the service doesn't; the other is a queue that exists because demand outruns even a well-funded supply.

The workforce is the binding constraint

Topic 03 showed equipment density tracking national income. The workforce gradient is steeper and slower to fix, because a radiologist is not a capital purchase. It takes a decade to train one and a plane ticket to lose one: South Africa trains about twenty a year — replacement level — while roughly a third of Australia's foreign radiologists are South African imports. This is why hardware-first donor programmes so often produce idle machines. A scanner without a reader is a very expensive way to generate images nobody interprets, and it is the reason any serious access strategy has to solve for people, teleradiology and task-shifting before, or alongside, procurement.

Equity doesn't stop at the border

Even where imaging is plentiful, the benefit lands unevenly. US mammography cut breast-cancer mortality about 40% since 1990 — and yet Black women die of breast cancer at a 41% higher rate than White women despite lower incidence, are half as likely to receive same-day diagnostic follow-up after an abnormal screen, and in representative geographic analyses live in tracts with 68% fewer mammography units. Rural populations are 41% of a representative state's people but reach 22% of its excellence centres. The global access map, in other words, reproduces itself in miniature inside wealthy countries: the same statistic that celebrates a 40% mortality fall also indicts a 40% racial gap in who received it.

The cheapest intervention no one funds

What makes the imaging gap unusual is how favourable the economics of closing it are. The Lancet Oncology Commission's Harvard microsimulation put the cost of scaling up cancer imaging across all LMICs at $6.84 billion over 2020–30 — trivial by health-system standards — against a modelled lifetime productivity return of $179 per dollar and 2.46 million cancer deaths averted. Even on the most conservative human-capital accounting the return stays above $30 per dollar. Numbers like these are why, in May 2025, the World Health Assembly adopted its first-ever resolution on medical imaging capacity. The evidence that closing the gap pays for itself has existed for years; the missing ingredient was political priority, and about 80% of cancer's lost life-years sit in countries receiving about 5% of its funding. The gap is a choice, not a constraint.

On the data. Radiologist-density figures come from different sources, definitions and years (2020–2026) and should be read as orders of magnitude, not precise counts; the "<2 per million" low-income and "97.9 per million" high-income figures are Lancet Commission / secondary-analysis estimates. Canadian wait times are physician-survey medians (13.1% response rate), not administrative records, and the Fraser Institute is a think-tank with stated policy positions — the underlying CIHI data confirm the direction if not every figure. US disparity statistics are observational and reflect many causes beyond imaging access (biology, treatment, insurance). The $179-per-dollar return and all death-averted figures are modelled projections from 2020, not realised results.

Sources

  1. Medical imaging and nuclear medicine: a Lancet Oncology Commission (2021) — workforce gap, deaths averted, ROI
  2. Lancet Oncology Commission — PubMed record ($6.84B cost; $179.19 per $1; 2.46M deaths averted)
  3. WHO — 78th World Health Assembly, Resolution WHA78.13 "Strengthening Medical Imaging Capacity" (24 May 2025)
  4. IAEA — WHA resolution on strengthening medical imaging capacity (2025)
  5. RSNA — WHO resolution on strengthening imaging capacity
  6. EU-REST study, Insights into Imaging 2025 — EU radiologist density 127/M (range 51/M Bulgaria to 270/M Sweden)
  7. How many radiologists are there in the world? (2025) — global distribution, ~45/M mean, income-group densities
  8. Density and distribution of radiologists in Ghana, Chin J Acad Radiol 2023 (Ghana 3/M; Nigeria figures)
  9. RSNA Daily Bulletin — Malawi (1:8.8M) and Tunisia (1:23,000) workforce ratios
  10. South Africa radiologist shortage (RSSA / World Health Expo 2024) — ~1 per 100,000; ~20 trained/yr; brain drain
  11. RANZCR 2020 Clinical Radiology Workforce Census (Australia) — overseas-trained graduates: India 7.8%, South Africa 5.7%, UK 5.0%
  12. Fraser Institute — Waiting Your Turn 2025 (MRI 18.1 wk, CT 8.8 wk, PEI MRI 52 wk)
  13. CIHI — diagnostic imaging wait times persist above pre-pandemic levels (2025)
  14. Breast cancer disparity in underserved women, RadioGraphics 2023 (SEER: +40.7% Black mortality)
  15. Emergence of the racial disparity in US breast-cancer mortality, NEJM 2024 (~40% mortality fall since 1990)
  16. Racial disparities in same-day breast diagnostics (2025) — Black women ~50% less likely to get same-day follow-up
  17. Delaware mammography catchment & location-allocation analysis, PMC 2023 (rural 41% pop / 22% BICOEs)